Researchers have introduced Aurora-X, a large-scale time series foundation model designed to overcome limitations in training potential and architectural versatility. The model employs a progressive curriculum, starting with channel-independent pretraining and advancing to incorporate cross-variable dependencies and future covariates. Aurora-X features a novel pattern-guided mixture-of-experts for efficient capacity expansion and an implicit quantile network head for flexible probabilistic forecasting. Experiments across multiple benchmarks demonstrate state-of-the-art performance compared to existing models. AI
IMPACT Advances time series forecasting capabilities with a versatile and scalable foundation model.
RANK_REASON Research paper detailing a new model architecture and its performance. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- DAG-Bench
- DagsHub
- FEV-Bench
- GIFT-Eval
- Gotit.pub
- Hugging Face
- IArxiv
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
